Weight estimation of sows during pregnancy using machine vision

Weight estimation of sows during pregnancy using machine vision

The body condition of sows, before and during pregnancy, has been shown to affect the quality and survivability of their offspring. Therefore, it is desirable to monitor and manage the body condition of sows to ensure that their nutritional and environmental requirements are met. A sow’s condition can be determined from its body shape, back-fat measurement and weight relative to its age and parity. However, often these factors are not monitored frequently due to the high level of labour required. As weight is a strong indicator of condition, a machine vision system was developed to continuously estimate, without operator involvement, the weights of pigs from their size. This article presents the weight estimation results achieved by the system for sows between days 71 and 82 of pregnancy, and a brief insight into the morphological changes in two sows’ body dimensions determined by the system which occurred in the days before and after giving birth. The system determined the average weight of the group of pregnant sows to within 1.5% mean-relative error of the actual group average weight. Using the system’s combined dimension and shape filtering method, 82% of the individual weight assessments of these sows were found to be within 5 kg of their actual weight. The system identified clear changes in body dimensions for two sows before and after giving birth. The effect that pregnancy has on the shape and body dimensions of sows should be investigated further as it is likely to be a contributing factor to poorer weight estimates. However changes in body dimensions may also be used to model and classify condition or the onset of pregnancy.

Files associated with this item cannot be displayed due to copyright restrictions.